This week saw launches of neat demos and tutorials, updates to dark theme toggles, powerful open-source cloud icons for PlantUML, and killer enhancements to GitHub user data tools. Also, the IITM Tools in Data Science course got crucial fixes, while an AWS-AI combo added smarter answers with citations.
New tools and major enhancements make data handling and GitHub interaction smoother and friendlier.
- GitHub User Data Extractor revamped: Now you can input usernames with or without '@', full URLs, and GitHub Pages links. It neatly formats dates, numbers, adds clickable links, and displays avatars as images (b1b5ed4). Yes, you really needed another GitHub abbr, but this one’s practical.
- JSON to CSV converter improved: Auto-detects JSON type (array or object) and preserves key order in output for more intuitive results (afbaef6). No more scrambling your columns!
- GitHubStars tool gets example input: Paste example markdown links and see live updates on star counts right away (7aea2b6).
- UI tweaks for larger demo grid: Added support for 4 columns on extra-large screens to avoid lonely widows in your demos gallery (0c2c0de).
- Preparing for broader repo search in podcast tool: Created a branch to accept GitHub repo search queries, a superset of just usernames (see related PR).
If PlantUML diagrams have been a chore, not anymore.
- New PlantUML architecture diagram tutorial: How to guide ChatGPT to create cloud architecture diagrams using official AWS, Azure, and GCP icons. Includes a handy bash script to generate those icon includes (3b22f54, d9437f8). Now you’ll have cloud diagrams that actually look like clouds, not just boxes.
- Improved prompt and instructions: The README now carefully guides you step-by-step, plus fixes minor errors to keep you on track.
Your dark theme toggler got smarter and safer.
- Treat unknown themes as ‘auto’: If someone tries an unsupported theme value, it falls back to using the system preference instead of showing an error (f506d4cb). Because sometimes you don’t want humans typing random theme names.
- Added tests: Automated tests in Vitest verify theme behavior across light, dark, auto, and invalid values (dark-theme.test.js).
- New example HTML: A fully working standalone demo page with Bootstrap 5.3 and dropdown button to toggle themes is included (dark-theme.html).
The online Data Science course materials remain polished and up to date.
- Corrected relative links in 2025 Jan content: Updated many file and sidebar URLs to use correct paths for the May 2025 semester project 1 notes release (d186af1). Less broken links, more learning!
- Moved virtual TA project images into ‘images/’ directory: Keeps the repository tidier and references consistent (de969a3).
- Removed confusing note about GA3 preponement: The graded assignment 3 submission date stays as June 1, 2025 (c04fb74).
- Minor typo fix: Fixes an important sentence about GitHub Pages on the course site (13d50dd).
New repo alert! A fresh start tracking scholarly topic trends with AI help.
- Repository created and initial commit: This app uses Large Language Models to interpret data trends in simple English (55c0ad6).
- Interpretation panel added: Now you can ask the LLM to summarize and explain topic trends from arXiv data with a customizable prompt. The output is neatly rendered with markdown (55c0ad6).
- User interface improvements: Prompt text area expanded, default prompt filled, and interpretation UI toggles on demand. Slides your papers into plain English territory in style.
Fresh setup to generate synthetic personas for marketing surveys and testing.
- New repository created: This tool uses LLMs to automatically create market research audience personas (create event). Because who has time to make up personas anymore? Let the AI do the imagining.
Your natural language API querying tool keeps growing smarter and more integrated.
- Added Google Analytics API support: Query web traffic and user behavior data directly via natural language (fe779f89). Great for getting quick insights from your analytics without diving into the dashboard.
- Integrated OpenAlex alongside Crossref API: Query scholarly works via multiple APIs transparently in one chat experience (0fae636). If you’re serious about academic data, this makes it easier.
- Added status display of current fetches: When the app talks to an API, you see which page it’s fetching, boosting transparency and user feedback (0fae636).
- Graceful error reporting in LLM calls: Instead of crashing, errors now display in the chat flow helping you debug or refine your query (7bf16c15).
Sharpened the retrieval augmented generation (RAG) on AWS with better answers and full stack.
- New OpenSearch indexing Python scripts: Added powerful tools to chunk, index, and embed Markdown or text docs for vector search on OpenSearch (index.py, c05659e). Because your vector DB needs a strong librarian.
- Hybrid search with query rewriting and embeddings: Uses GPT-4 to rewrite queries and craft sub-questions, then runs hybrid BM25+k-NN search (search.py, 99b5d31). Makes your searches smarter, like a bit of sleight-of-hand behind the scenes.
- Add answer generation with citations: Retrieves relevant documents and prompts GPT to answer based on them, including citations with cross-references for trustworthiness (9dc1cf0). Finally, an AI that can say “I got this from source #3.”
- Terraform plus Docker for deployment: Infra now includes OpenSearch domain with secure basic auth, builds container image with Docker, managing via tofu CLI (main.tf, dd2c19d). Deploy with confidence on AWS.
Your favorite Generative AI WhatsApp group’s podcast automation just got tighter on date alignment.
- Correct podcast week range: The podcast folders now correspond exactly to the Sunday they’re released and cover the full prior week to Saturday, avoiding future data leakage (fe01430). Half the mystery of “Did this episode cover last week or this week?” solved.
- Bundle the PlantUML tutorials with your cloud diagrams for easier shareable architecture docs.
- Try out the enhanced GitHub User Data Extractor for packing rich user profiles effortlessly.
- Explore integrating the AWS RAG setup with your own document datasets — citations on demand!
- Experiment with the Topic Trends app and feed it some arXiv data for research insights.
- If you’re building UI themes, steal a look at the Bootstrap-dark-theme’s neat fallback logic and tests.
- Join the Tools in Data Science public course or use it for teaching future data pros.
- And maybe… build your own tiny API agent with that new Google Analytics integration?